Protecting Networks from Large-Scale Physical Attacks and Disasters

Telecommunication networks play a vital role in the day-to-day routine of all sectors of our society. During a crisis, telecommunications is essential to facilitate the control of physically remote agents, provides connections between emergency response personnel, and eventually enables reconstitution of societal functions. However, telecommunication networks heavily rely on physical infrastructures (such as optical fibers, amplifiers, routers, and switches), and therefore, are vulnerable to natural disasters, such as earthquakes or floods, as well as to physical attacks, such as an Electromagnetic Pulse (EMP) attack. Increasingly, networks use a shared infrastructure to carry voice, data, and video simultaneously; hence, failures in the physical infrastructure will lead to a break down of vital services.

Physical attacks or disasters affect a specific geographical area and will result in failures of neighboring components. Therefore, it is crucial to consider the effect of attacks on the physical (fiber) layer as well as on the (logical) network layer. Although there has been a significant amount of work on network survivability, most previous work considered a small number of isolated failures. In contrast, this project considers events that cause a large number of failures in a specific geographical region.

Disasters such as EMP attacks, earthquakes and foods affect a specific geographical area and can significantly disrupt the Internet service over the affected area. We studied the robustness of the physical fiber-optic network to such a regional disaster. To that end, we developed an algorithm that identifies the part of the network that should be fortified to reduce the impact of a disaster or attack, and can be used for making the network more robust. Based on our understanding from this study, we addressed the problem of providing a protection guarantee to such correlated failures. To deal with the correlated failures more systematically, we proposed a new Probabilistic Shared Risk Link Group (PSRLG) failure model that can effectively describe geographically correlated failures. The PSRLG model enabled us to systematically study path protection problems under correlated failures. Using this model, we developed path protection schemes that find a pair of paths with the minimum joint failure probability under correlated failures. Our new model and results will have a significant impact in modeling, understanding and building communication networks robust to natural disasters or intentional attacks.

The main goals and accomplishments of this project include:
(*) Development of techniques to identify the most vulnerable parts of the network.
(*) Development of tools to evaluate network reliability in the face of geographically correlated failures.
(*) Develop network designs that are robust to geographically correlated failures
(*) Develop dynamic restoration algorithms that will improve the resilience of networks to geographically correlated attacks and prevent cascading failures.